practice today,
type better tomorrow

This test starts the moment you do. There is no countdown to sit through: the clock begins on your first keystroke, so you can settle in, read the first few words and start when you are ready. Words stream in continuously rather than being fixed in advance, which means you cannot run out of text however fast you type, and no run ends early because you got further than expected. Wrong characters are marked as you go, and Backspace works normally if you want to correct them.

The passage is built from the two hundred most common words in English, shuffled fresh for every attempt. That is a deliberate choice rather than a shortcut. Common words are the letter sequences your hands are most likely to have automated, so the test measures your typing rather than your vocabulary or your luck with an unusual passage. It also keeps repeat attempts comparable, which matters more than it sounds once you start tracking progress over weeks.

You can run the test at six fixed lengths — 30, 60, 120, 180, 240 and 300 seconds — or choose free run, which has no timer at all and finishes when you press Enter. Shorter runs are useful for a warm-up; longer ones give a figure you can actually reproduce. When the clock stops you get five numbers, a graph of your speed sampled once per second, and a character-by-character review showing exactly where you diverged from the passage.

The five-character word is inherited from typewriter-era testing, and it survives because it solves a real problem: counting actual words would reward whoever happened to be handed a passage of short ones. Normalizing to five characters takes vocabulary out of the measurement, which is what lets a score from one test be compared against a score from another. It also means your WPM is really a character count in disguise, worth remembering when a passage of long words feels harder than the number suggests.

One behavior surprises people on their first attempt. The moment you start typing, everything that is not the passage fades out, from the header and footer to the controls and captions, and it all comes back when you stop. Moving the mouse mid-run pauses the test rather than letting the clock keep running, because reaching for the mouse almost always means you have stopped typing. Clicking the passage again resumes from where you left off.

Nothing is required to start. There is no sign-up, no installation and no advertising interrupting a run, and every result is saved automatically so your history builds up whether or not you ever create an account. If you want the full walkthrough of every control and setting, the complete guide to using the site covers it in detail.

Start with the length. Use sixty seconds at minimum for any figure you intend to quote, and three minutes or longer for the most honest one. A thirty-second run is genuinely useful as a warm-up, but it can be carried by a lucky passage and a burst of familiar words, and it ends before concentration slips, hands tire or an awkward letter sequence turns up. Expect a thirty-second score to sit roughly ten percent above your sustained speed.

Then discard your first attempt of any session. It is a warm-up whether you intended it as one or not — almost everyone types measurably slower on the first run than the second, and comparing a cold first attempt from last week against a warm fifth attempt today is the most common way people convince themselves they are improving faster or slower than they really are.

Take several runs across different days rather than several in one sitting, and use the median rather than your best. Your best is a ceiling you touched once under favorable conditions; the median across a handful of attempts is the speed you actually own, and it is the figure that will hold up if someone asks you to reproduce it. Keep the duration fixed while you do this, because scores from different lengths are not interchangeable.

Take the reading under conditions you can repeat. Test on the keyboard you use every day rather than an unfamiliar one, sit the way you normally sit, and avoid measuring at the end of a long working day when your hands are already tired. None of that is about producing a flattering number. It is about removing the variables that would otherwise make two of your own results incomparable.

Then record the figure, or let the site do it for you. Every run is saved automatically, and your history groups results by duration with a running average alongside your best, which is the comparison that actually shows progress. A single score tells you where you were on one afternoon; a month of medians tells you whether anything is genuinely changing.

If you want the cold-first-run rule without wasting a real attempt, warm up in free run instead. It has no timer, so you can type for thirty seconds or two minutes purely to get your hands moving, then switch to a timed duration for the measurement that counts. That way your first recorded run of the session is already a warm one and you are not throwing a result away to get there.

Finally, record accuracy and consistency alongside the headline number, not instead of it. A speed figure with no accuracy attached is not comparable to anyone else's result, including your own from last month. Once you have a reliable figure, average typing speed sets out the bands and shows where your score actually places you.

Five numbers appear when a run finishes. Net WPM is the headline figure and the one the leaderboard ranks: it counts only characters that matched the passage. Raw WPM applies the identical formula to everything you typed, right or wrong, so it reflects how fast your hands actually moved. The gap between those two is the most diagnostic thing on the screen, because it is the cost of your mistakes expressed in words per minute.

CPM is characters per minute — the same correct-character count behind your net WPM without the division by five. It carries no information your WPM does not; it is the same measurement in different units. Accuracy is the share of keystrokes you got right out of every key you pressed, counted live as you type and including corrections. It is highlighted on the results screen once it reaches 97 percent, which is roughly the point where errors stop being the thing holding your score down.

Consistency is the one most people have not seen before. Your speed is sampled once per second across the whole run, and consistency reports how much those readings varied relative to your own average, on a scale from 0 to 100. Metronomic typing approaches 100; a run made of bursts between pauses can score below 60 even at a strong WPM. It is a measure of how automatic your typing has become rather than how fast it peaked, which is why it tends to improve before your speed does.

You will occasionally meet the same measurement under other names. Keystrokes per minute is identical to CPM when only correct keys are counted, and keystrokes per hour, the unit data entry roles usually quote, is that figure multiplied by sixty. To convert KPH into words per minute, divide by sixty and then by five, which puts 10,000 KPH at roughly 33 WPM.

One thing worth clearing up: a word here is not a word. Because the count is characters divided by five, a passage of three-letter words and a passage of ten-letter words should score you about the same for the same amount of physical typing. That is precisely the point of the convention, and it is why WPM measures keystrokes rather than fluency with language.

Read the five together rather than as five grades. A wide raw-to-net gap points at errors; a low consistency with a decent average points at uneven familiarity; a clean, narrow gap at a low number points at genuinely slow hands. The exact arithmetic behind every figure, including why CPM and WPM are not always exact multiples on screen, is worked through in typing scores explained.

Three things about this test push toward a reliable reading. Timing begins on your first keystroke rather than on page load, so no part of your score is lost to a slow start. Speed is sampled every second for the whole run rather than computed once at the end, which is what makes the graph and the consistency score possible. And accuracy is counted per keystroke at the moment each key is pressed, so it reflects how cleanly you typed rather than how carefully you went back and tidied up afterwards.

Three things limit any single result, here or anywhere. Test length matters most: short runs scatter, longer ones converge, and thirty seconds is not a measurement so much as a sample. Passage difficulty matters next — this test uses common English words without punctuation or capitals, which keeps repeat attempts comparable but also means your score here will usually sit above what you would produce on a passage full of proper nouns, numbers and symbols. And the keyboard in front of you is worth a few words per minute either way until you have adapted to it.

It is also worth knowing what a typing test cannot tell you. It measures transcription — copying text you can see — so it says nothing about how fast you compose your own sentences, which for most people is bounded by thinking rather than typing. It will not predict your speed on code, on numeric data entry, or across an hour of sustained work. Those are different tasks with different bottlenecks, and a single figure from a one-minute run does not reach them.

The per-second sampling is what makes the graph honest rather than decorative. Instead of dividing a total by a duration once at the end, the test records where you actually were at every second of the run. A 60 WPM score built from a steady pace and a 60 WPM score built from bursts between stalls look completely different on screen even though the headline figures match.

It is also scored more forgivingly than a workplace assessment, and that gap catches people out. Employer tests commonly subtract each uncorrected error as a whole word from your total and set a hard accuracy floor beneath which a run simply fails. Here an incorrect character contributes nothing but does not deduct. Expect a formal assessment to come in below your score here, and see typing speed for jobs for how that arithmetic works.

The practical answer, then: treat one run as indicative and the median of several as accurate. Take them at the same length, on the keyboard you actually use, discarding the first of each session. Do that and the number will be stable enough to track over months, which is the only thing a typing test is really good for.

Two conditions have to hold before any comparison is meaningful. The first is accuracy: a 60 WPM run at 88 percent accuracy is not a 60 WPM result in any useful sense, because the corrections you did not make are still sitting in the text. Treat 95 percent as a floor and 97 or better as the level worth working toward. The second is duration. The same typist reliably scores higher on a shorter test, so comparing your thirty-second best against someone else's five-minute figure tells you nothing at all.

Percentile claims are worth treating with more suspicion than most sites encourage. Every widely circulated typing percentile table draws its data from people who chose to take an online typing test, and that group is not the general population — typing tests attract people who care about typing, which skews the sample upward. Neither that bias nor the smaller counter-bias of people who suspect they are slow can be measured, so the resulting percentiles cannot be corrected.

What can be said confidently is directional. The distribution has a long right tail: speed cannot fall below zero but can climb past 120, which pulls the average above the score most people actually get. That is why comparing yourself to an average flatters roughly half of everyone and discourages the rest, and why the band immediately above yours is a far better target than any published midpoint.

Pair any figure with an accuracy number before judging it. Somebody typing 55 WPM at 98 percent is in every practical sense faster than somebody typing 65 at 90, because only correct characters count and mistakes are paid for twice, once to make and once to fix. A speed with no accuracy attached is a number rather than a skill, which is why both are shown together here.

Age is worth setting aside too. Charts breaking typing speed down by age circulate widely and almost none state their methodology or sample size. What is better supported is that keyboard exposure explains far more of the variation than age does. Somebody who has written at a keyboard daily for twenty years will usually outpace somebody half their age who has spent most of that time on a phone.

For a real comparison rather than an estimated one, the leaderboard shows where a given score places among everyone ranked at the same test length — which is the only fair way to do it. For what each band actually means and what separates 40 from 60 from 80, average typing speed works through the full distribution.

As a guest you get the whole test: all six timed durations plus free run, all five metrics, the per-second speed graph, the character-by-character review of what you typed, and a saved history of your runs with your personal best tracked per duration. Nothing is degraded or watermarked. The one thing a guest cannot do is appear on the public leaderboard, and that is a deliberate design decision rather than an upsell — an unranked anonymous entry would make the board meaningless.

Signing in with Google is optional and adds three things. Your history moves from living in one browser to living in an account, so it survives clearing your browsing data and follows you to other devices. Your personal best for each timed duration becomes eligible for the leaderboard. And the history you built as a guest carries over rather than being stranded — signing in attaches it to your account instead of starting you over.

There is nothing to install either. The test runs in the browser you already have, on desktop or mobile, with no extension, no download and no permissions to grant. It works offline once the page has loaded, since the test itself does not need the network to measure anything — only saving to an account does.

Free also means free of the usual friction. There is no run limit, no daily cap, no waiting between attempts and no interstitial screen before your result appears. You can take the test fifty times in a row if you want to, and the fiftieth behaves exactly like the first.

The leaderboard is included rather than gated behind a premium tier, and so are the history, the graph and the character review. If a feature is described anywhere on this site it is available to everyone. The only thing signing in changes is where your history is kept and whether your best score becomes eligible for ranking.

Not creating an account also means there is nothing to unsubscribe from later. No email address is requested at any point as a guest, so there is no confirmation step, no verification link and no mailing list attached to taking a test. If you do sign in, it is with an existing Google account rather than a new set of credentials to remember.

Two things the site deliberately does not do are worth stating plainly, because both are common elsewhere. It publishes no ratings or review scores, so there are no invented star ratings anywhere on it. And it does not issue certificates, which means if a job application requires one from a specific named body, this is not where to get it. What you can get is an accurate measurement, free, without an account.

The trade-off with guest storage is that the records live in that browser on that device. Clearing your browsing data removes them, and they will not follow you to a different browser or a different computer. Nothing from a guest session is ever published anywhere — guest results cannot reach the leaderboard by design, so a run you took without signing in stays entirely between you and your machine.

Signing in changes where the history lives rather than what is collected. Your runs are kept against your account so they survive a cleared cache and are available wherever you sign in, and your display name and profile photo come across from your Google account so the leaderboard can show who a score belongs to. Whatever you built as a guest beforehand is attached to the account rather than discarded.

Only one thing is ever made public, and only once you are signed in: your single best result for each timed duration. The individual runs behind it are not published, free runs are never ranked, and there is no public profile listing your history. Your own history page is kept out of search results entirely, because it shows your data and there is nothing on it for anyone else to find.

The fifty-run cap exists so the history stays useful rather than becoming an archive nobody reads. Fifty attempts is comfortably enough to show a trend across several weeks of regular practice, and your personal best for each duration is tracked separately, so a strong run is not lost simply because you have taken a lot of tests since.

Across multiple devices the behavior follows from where the records live. Signed in, your history is the same wherever you sign in. As a guest, each browser keeps its own separate set, so a run on your laptop will not appear on your phone and neither will know about the other. Signing in is the only thing that joins them up.

Sharing is the one case where a result leaves your own view, and it is always something you choose rather than something that happens automatically. Sharing a run mints a link to that single result, showing its score, duration and graph and nothing else about you beyond the display name on your account. Those pages are excluded from search indexes, so a shared link reaches whoever you send it to and nobody else.

You can also delete everything. Clear history on your history page removes every saved run, on the device and in your account if you are signed in, and it cannot be undone. What is stored, for how long, and which third parties are involved is set out in full in the privacy policy.

Test length is the largest single factor. Short runs systematically flatter you: a thirty-second test can be carried by a burst of familiar words and it finishes before fatigue, drift or an unfamiliar sequence has a chance to appear. Longer runs let all of that surface, so scores fall and then stabilize — usually by around the two-minute mark. If two of your results disagree sharply, check whether they came from the same duration before looking for any other explanation.

Warm-up is next, and it is the one people most often forget to control for. The first run of any session is measurably slower than the second for almost everyone, so a cold attempt and a warm attempt are not the same measurement. Passage familiarity matters too: the words are shuffled fresh each time, so one run may happen to hand you sequences your hands already know while the next does not.

Then a set of smaller, consistent influences. An unfamiliar keyboard costs most people a few words per minute until they adapt to its travel and spacing. Fatigue shows up on longer runs and barely at all on short ones. Posture affects the reading through your hands rather than your motivation — wrists pressed flat to the desk anchor the hand and force the fingers to stretch. And attention, unsurprisingly, is worth several WPM either way.

The graph is the fastest way to see variation inside a single run rather than between runs. A line that climbs and then settles means you were warming up. Repeated sharp dips mean specific words are costing you, and the character review underneath will name the keys involved. A jagged line with a strong average is the signature of speed carried by familiar words.

Time of day matters more than most people expect, and in both directions. Typists tend to be slower first thing and slower again late in the evening, with a broad plateau in between. If you are trying to detect a real improvement of a few words per minute, measuring at roughly the same time of day removes one more source of noise from the comparison.

One more difference is worth allowing for if you compare a score here against one from somewhere else. The passage here uses common words without punctuation or capital letters, which keeps repeat attempts comparable but also removes the two things that slow most typists down. A test with heavy punctuation, proper nouns or numbers will usually return a lower figure from identical hands, and that is the test being harder rather than you being slower.

The fix is procedural rather than technical. Fix one duration, take the run cold as the first of a session, use the same keyboard, and compare the median of several attempts across different days rather than any single score. Do that and the remaining variation is small enough that a genuine improvement becomes visible instead of being lost in the noise — the practice side of which is covered in drills and a practice routine.

The arithmetic is straightforward once you know what is being counted. Press 300 keys, get 12 of them wrong, and accuracy reads 96 percent, because 288 of 300 were correct — and the Backspace presses you used to fix those 12 do not remove them from the denominator. This is why a fast typist who corrects heavily can end up with a lower accuracy figure than a slower one who simply made fewer mistakes.

That design is deliberate rather than an oversight. Accuracy is meant to measure how cleanly you type, not how diligently you proofread. Someone who hits the right key first time is doing something genuinely different from someone who types twice as many keys and corrects the difference, and a metric that scored both at 100 percent would be hiding the distinction that matters most if you want to get faster.

There is a useful diagnosis hiding in the number. If accuracy is low while the finished text looks clean, you are not a careless typist so much as a fast one working above the pace your hands have actually learned. Deliberately typing about ten words per minute below your maximum usually recovers most of it within a week, and because only correct characters count toward your score, the net effect is that you get faster by slowing down.

Not every test counts it this way, which is why your accuracy figure can shift when you switch sites. Some score accuracy against the finished passage rather than against your keystrokes, which produces a far more forgiving number. A heavily corrected run can read as 100 percent there and 94 percent here from identical typing. Neither is wrong; they are answering different questions.

The 97 percent threshold is highlighted for a reason. Below it, errors are usually the single largest cost in your score, and reducing them raises your net speed faster than any amount of pushing for pace. Above it, accuracy has stopped being the constraint and speed work becomes the better use of your time. It is a switching point rather than a grade.

It is also worth separating accuracy from consistency, because a low reading on one says nothing about the other. Accuracy asks how many keys you got right; consistency asks how evenly your pace held. You can be metronomic and inaccurate, or erratic and near-perfect, and the two point at different fixes: accuracy usually responds to slowing down, while consistency responds to practicing the specific words that make you stall.

Look at which characters failed rather than how many. Almost everyone has two or three responsible for a disproportionate share of their errors, and the character-by-character review under your results names them. Drilling those in isolation fixes in days what general practice takes weeks to erode. The habits behind the most common patterns are in common typing mistakes, and the formula itself is in typing scores explained.

Whether the resulting score means much is a different question, and the honest answer is that it measures something else. Touch typing is a ten-finger skill built on fixed hand positions and tactile feedback from physical keys, and none of that exists on a glass screen. A phone score reflects thumb-typing speed, which is a real and useful ability but not the one a words-per-minute figure is conventionally reporting.

So do not compare the two. A phone result and a desktop result are not interchangeable, and treating them as the same metric will make your history look erratic for no reason. If you track progress on mobile, track it against your own mobile results, and keep desktop attempts as a separate line entirely. The per-duration filters on your history page make that easy enough to eyeball.

Pairing an external keyboard with a tablet changes the picture completely. Once you have physical keys under ten fingers, the measurement becomes directly comparable to a desktop run, and a tablet with a keyboard case is a perfectly reasonable place to practice. The variable that matters is the keyboard, not the size of the screen or the operating system behind it.

Predictive text and autocorrect are the other complication on mobile. Depending on your keyboard settings the device may be completing or silently altering words as you type, which either inflates your speed or produces errors you did not actually make. If you want a phone score that means anything, turn both off for the duration of the run.

There is a physical ceiling worth knowing about as well. Two thumbs on a glass surface cannot approach what ten fingers on physical keys can do, and thumb-typing speeds top out well below competent touch typing. That is not a limitation of the test or of you. It is the input method, and it is exactly why comparing the two is not meaningful.

None of which makes mobile practice pointless. Taking a test on a phone still builds familiarity with where letters sit, still gives you a record you can compare against your own earlier mobile runs, and is a perfectly good way to keep a daily habit going on days you are away from a desk. Just treat the number as its own separate measurement rather than as your typing speed.

One practical note for mobile: the test needs the passage to hold keyboard focus before your keystrokes reach it, so tap the passage once to bring up the on-screen keyboard and start the run. If typing appears to do nothing, that focus is what is missing. Everything else — durations, free run, saving, sharing — works identically to desktop, and the full guide covers each control if you want the detail.

The consistency score is the clearest example. Your speed is sampled every second for the whole run, and consistency reports how much those readings varied relative to your own average. It matters because two typists can average the same WPM while one holds a steady pace and the other fires off bursts between stalls — and the steady one has automated the movement while the other has not. Consistency detects that directly, and because it improves before speed does, it is usually the earliest honest evidence that practice is working.

Every timed run also produces a graph of that per-second series, with your personal best for the duration drawn in as a reference line, plus a character-by-character review showing exactly where you diverged from the passage. Together those answer the question a bare score cannot: not how fast you typed, but where you lost the time and which keys cost it.

The measurement is also kept deliberately honest. Accuracy is counted per keystroke rather than against the tidied-up text, so corrections are visible in the number instead of being hidden by them. Net and raw speed are both reported, so the cost of your errors is explicit. No ratings or review scores are published anywhere on the site, because none have been collected — and no scores are inflated to make anyone feel faster than they are.

Free run is a small feature that turns out to matter. Removing the clock entirely changes how you can practice: you can work on a specific set of keys, warm up properly, or simply type without a countdown in your peripheral vision. It records the same five metrics as any other run, and it is deliberately never ranked, since a score from an unbounded test cannot be compared fairly against a fixed one.

The leaderboard is six independent boards rather than one, and there is a reason behind that. Because the same typist reliably scores higher over thirty seconds than over five minutes, a single combined ranking would mostly sort people by which duration they picked. Keeping them separate means a 300-second entry competes against other 300-second entries, which is the only version of the comparison that says anything.

A lot of the difference is simply friction that is not there. The timer starts on your first keystroke instead of counting you down, Tab restarts a run instantly if you want to abandon it, Enter takes you straight into the next one, and no screen appears between finishing and seeing your result. Individually those are small; together they are the difference between taking one test and taking ten.

Around all that: six timed durations from 30 to 300 seconds plus an untimed free run, results saved automatically without an account, a live leaderboard with a separate board per duration, shareable result links, a dark theme, an optional keystroke sound, and no advertising interrupting a run. If you want to see how each of those works before trying them, the complete guide walks through the lot.